INDEX

Explanations

references to political figures who are currently in office

oai_token-act-pair · gpt-3.5-turbo

instances of the word "incumbent" in various contexts

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

 Deaths

-0.74

Shape

-0.68

 Style

-0.68

hed

-0.66

sight

-0.64

hex

-0.63

EMA

-0.62

helm

-0.61

FO

-0.60

 Slip

-0.59

POSITIVE LOGITS

 incumbent

4.13

 incumb

2.75

 challenger

1.34

 frontrunner

1.15

 entrenched

1.11

 reigning

1.10

ocrat

1.08

 challengers

1.08

 outgoing

1.05

establishment

1.02

Activations Density 0.011%

references to political figures who are currently in office

instances of the word "incumbent" in various contexts

No Known Activations